This is mostly true, unless you specialize in legacy technology to serve a niche, like COBOL mainframes. In this regard the profession is similar to that of a doctor or a lawyer, where the knowledge body is not as impermanent.
I think the learning mechanic still holds. You've still got to learn the old paradigms, but if/when there's an impetus to migrate, you need to learn the new paradigms and (hopefully) provide useful & cost-effective knowledge during the migration (my last project as an intern at a Fortune 500 company was to rewrite a legacy batch-processed mainframe Cobol application to a web-based .Net application. I had to learn at least enough Cobol to understand what was being done in order to replace it. Went from about a 45kloc Cobol program to about a 10k .Net program w/ about 5k of xml configuration. Point is: even if you're not going to write in a language, sometimes you need to understand a different one. I think I saw perhaps the worst language I've ever seen: "MOVE NEXT SENTANCE", which jumps, unconditionally to the next period in the source. Yes, a statement "MOVE NEXT STATEMENT" jumps to the very next period, aka ".". It's a very large goto statement that jumps to an extremely terse label. I don't mind appropriate usage of gotos in C/C++/C#, but this construct is just ripe for abuse and misunderstanding.
